Today, as multimedia contents become huge, there is a demand for an automated system that can select data desired by a user. Recently, due to development of a moving picture stream recording/reproducing apparatus, for example, personal video recorder (PVR), which records an incoming moving picture stream and reproduces the recorded moving picture stream, an automated system that can select data desired by a user is demanded very frequently.
Particularly, in the moving picture stream recording/reproducing apparatus such as PVR, the function of selecting and searching data desired by a user is very useful for an edition of a user's own contents as well as a simple browsing. In other words, users want to watch contents in which advertisements or uninteresting portions are excluded and want to edit and store desired portions separately.
In addition, according to the moving picture stream recording/reproducing apparatus such as PVR, a moving stream which goes on the air can be recorded on a digital media, and a replay of a corresponding content can be always requested. Further, it is possible to provide a random access to a desired location or a high-speed quick view function, which is impossible in an analog video.
Meanwhile, as one of methods for analyzing the multimedia contents, video index technologies have been researched for long terms. Among them, a shot segmentation technology has been known as a technology that analyzes the video contents at a very high accuracy. The shot segmentation technology divides the video contents into shots that are a physical edition unit. In general, most of the shot segmentation technologies have the high accuracy of 95% or more. Additionally, the shot segmentation technology can be integrated into the moving picture stream recording/reproducing apparatus such as PVR. For example, shot index can be generated through the recording and analysis of the video contents by using the shot segmentation technology.
A time-based skip is typically proposed for providing the random access to user in a TV-like environment, such as PVR. The time-based skip moves a reproduction location of the media 20 seconds after or 7 seconds before a current location, thereby moving to a location desired by the user. According to the time-based skip, however, it is impossible to accurately move to the desired location. Further, in order to move a similar location, a large number of user inputs are necessary.
In addition, fast forward/fast rewind (FF/FR) functions are provided to the user in order for the high speed searching. However, in case the FF or FR function is required for moving the desired location, the use must carefully watch the reproduction of the media in order to check whether the reproduction location arrives at the desired location while the media is being reproduced in FF or FR mode. Therefore, the user may suffer from dizziness during an operation of the FF or FR. Further, a difference between the user's recognition speed and the reproduction speed of the media makes it difficult to reproduce the media from the desired location.
In order to solve these problems, a scene-based skip has been proposed. The scene-based skip changes the reproduction location of the media by a scene unit, not by a shot unit. Here, the scene unit is a conversation structure unit that is greater than the shot unit. According to the scene-based skip, the reproduction location of the media is changed by the logical scene unit. However, in the application of a method for automatically detecting scenes, the current technology is limited to specific genre (news, drama, etc.). Therefore, it is difficult to apply the method to video contents of all genres. Further, since the user demands both the shot-based skip and the scene-based skip at the same time, it is difficult to provide a limited user interface (UI) with both functions.